| View previous topic :: View next topic |
| Author |
Message |
andrewwalker27
l33t
Joined: 27 Jun 2005
Posts: 660
|
 Posted: Tue Jun 30, 2020 8:06 pm Post subject: Ryzen experimental kernel risks |
 |
|
I'm interested in using the Ryzen kernel options but as they require the experimental use flag I'm a bit apprehensive, has anyone tried them out? I'm not using my machine for anything critical, and I'm already running ~amd64, but I'd like to know if they are reasonably stable to play with.
Anyone had any experience with them by any chance? |
|
| Back to top |
|
 |
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 45794
Location: 56N 3W
|
| Posted: Tue Jun 30, 2020 8:19 pm Post subject: |
|
|
andrewwalker27,
That's what backups, or snapshots are for.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
saboya
Guru
Joined: 28 Nov 2006
Posts: 518
Location: Brazil
|
| Posted: Tue Jun 30, 2020 8:55 pm Post subject: |
|
|
| I compile the kernel with AMD Zen 2 processor family. I've using this for several months with no issues. |
|
| Back to top |
|
|
kajzer
l33t
Joined: 27 Nov 2014
Posts: 776
|
| Posted: Wed Jul 01, 2020 1:48 pm Post subject: |
|
|
| It's safe, more important question is are there any benefits choosing for example znver2 as a processor family over "AMD/Athlon64..." |
|
| Back to top |
|
|
Anon-E-moose
Advocate
Joined: 23 May 2008
Posts: 4702
Location: Dallas area
|
| Posted: Wed Jul 01, 2020 2:05 pm Post subject: |
|
|
| kajzer wrote: | | It's safe, more important question is are there any benefits choosing for example znver2 as a processor family over "AMD/Athlon64..." |
For the kernel, I don't think it makes that much difference between generic x86_64 and znver2. For applications it can make a significant difference.
_________________
PRIME x570-pro, 3700x, RX 550 & 560 - 5.8 zen kernel
Acer E5-575 (laptop), i3-7100u - i965 - 5.5 zen kernel
---both---
gcc 9.3.0, profile 17.1 (no-pie & modified) amd64-no-multilib, eudev, openrc, openbox, palemoon |
|
| Back to top |
|
|
kajzer
l33t
Joined: 27 Nov 2014
Posts: 776
|
| Posted: Wed Jul 01, 2020 2:19 pm Post subject: |
|
|
| Anon-E-moose wrote: | | For the kernel, I don't think it makes that much difference between generic x86_64 and znver2. For applications it can make a significant difference. |
Sure, I think OP was thinking about the kernel.
I could be wrong though. |
|
| Back to top |
|
|
Goverp
l33t
Joined: 07 Mar 2007
Posts: 877
|
| Posted: Wed Jul 01, 2020 6:12 pm Post subject: |
|
|
I've used the hack to achieve a similar effect:
| Code: | | KCFLAGS="-march=native" make -j24 |
for some time. It still works on my desktop machine (which is znver2), but it's recently started failing to compile on my laptop (bdver4) box, fails with "Error 255 in kernel 5.7 compiling arch/x86/events/amd/ibs.o",.
For now I'm using the nearest standard equivalent. I expect the problem will be cured when there's a gcc update. It's possible the "experimental" patches do something more intelligent than simply "-march=native", but I've not tried.
_________________
Greybeard |
|
| Back to top |
|
|
kajzer
l33t
Joined: 27 Nov 2014
Posts: 776
|
| Posted: Wed Jul 01, 2020 10:47 pm Post subject: |
|
|
| Goverp wrote: | | It's possible the "experimental" patches do something more intelligent than simply "-march=native", but I've not tried. |
It's just "-march=native"
It's like that for every other cpu, I mean there's nothing more to it than what you would set in make.conf for 'march'
For znver2 it's just znver2 as well, etc.
You can check it anyway by extracting '/var/db/repos/gentoo/distfiles/genpatches-5.7-7.experimental.tar.xz' and looking at '5012_enable-cpu-optimizations-for-gcc91.patch'
Maybe I'm missing something  |
|
| Back to top |
|
|
Anon-E-moose
Advocate
Joined: 23 May 2008
Posts: 4702
Location: Dallas area
|
| Posted: Thu Jul 02, 2020 9:57 am Post subject: |
|
|
Sometimes march native works and sometimes it doesn't
For my oid fx cpu, it would crash and burn every time I used native (it was using piledriver)
it only worked when I set the cpu to bulldozer (previous gen) that worked fine, so I just left it like that, while I had that system installed.
Now, with the ryzen I still don't use native, I set it specifically for the cpu znver2 in this case.
_________________
PRIME x570-pro, 3700x, RX 550 & 560 - 5.8 zen kernel
Acer E5-575 (laptop), i3-7100u - i965 - 5.5 zen kernel
---both---
gcc 9.3.0, profile 17.1 (no-pie & modified) amd64-no-multilib, eudev, openrc, openbox, palemoon |
|
| Back to top |
|
|
Juippisi
Developer
Joined: 30 Sep 2005
Posts: 474
Location: /home
|
| Posted: Thu Jul 02, 2020 1:25 pm Post subject: Re: Ryzen experimental kernel risks |
|
|
| andrewwalker27 wrote: | I'm interested in using the Ryzen kernel options but as they require the experimental use flag I'm a bit apprehensive, has anyone tried them out? I'm not using my machine for anything critical, and I'm already running ~amd64, but I'd like to know if they are reasonably stable to play with.
Anyone had any experience with them by any chance? |
You can audit the experimental patches here:
https://dev.gentoo.org/~mpagano/genpatches/trunk/5.7/
They're the >=5000 ones. |
|
| Back to top |
|
|
Goverp
l33t
Joined: 07 Mar 2007
Posts: 877
|
| Posted: Fri Jul 03, 2020 8:24 am Post subject: |
|
|
Aha! Reading the optimization patch, I note a warning comment not to use "-march=native" with bdver4 cpus. That explains my problem, at least.
_________________
Greybeard |
|
| Back to top |
|
|
|
Kernel & Hardware
